sccomp_proportional_fold_change question

[irenerh]

Thank you so much for developing this great tool!

I’m analyzing compositional differences in a scRNAseq project using a design with interactions.
I have 3 different variables (genotype, tissue and treatment). I performed pairwise comparisons within one of the variables, for e.g. genotype, stratifying by combinations of two of the other predictors, e.g. tissue and treatment (formula_composition = ~ treatment*tissue*genotype).
The results from both sccomp_estimate() and sccomp_test() for all comparisons seem reasonable and good.
However, I’m trying to get the fold proportion differences for the different comparisons using sccomp_proportional_fold_change(), but it only allows to compare categories within one single variable.
Is there a way I can obtain the fold proportions for, e.g. genotype comparisons, subsetting for the different combinations of tissue and treatment?

Thanks!

[stemangiola]

Hello,

interesting. I have not really focused on interactions for sccomp_proportional_fold_change(). It could be a little project.

A quick workaround is, if all you interactions terms are categorical, is to create a new column concatenating them obtaining hardcoded interac tions.

Another attempt would be to try "A:B: X - A:C: X", as the actual covariate would have such a naming structure. Not sure if the current implementation would allow it.

I am sure, if someone wanted to contribute to the repository, it would be a fairly easy and fun mini-project, that would make them part of the sccomp family! ;)

[stemangiola]

Can. you please give us an example of how you would like to use

sccomp_proportional_fold_change()

With your model, please give us the preceding code you use, and possibly the anonymised cell metadata that works with your code.